3 ms·
Operant Conditioning in Generative AI Image Creation
- ilaksh 3y agoJust ask Stable Diffusion 1.5 to make an image of a human body (without a negative prompt). The first few times I tried that, I guess I experienced shock from the body horror I was being presented with. But eventually I learned that I just had to politely but specifically request that it NOT show me images of amputees, three-legged monsters, birth defects, conjoined twins, broken arms, extra fingers, etc. etc. Then it usually works pretty well.